JPH0589279A - Character recognizing device - Google Patents

Abstract

PURPOSE:To provide a character recognizing device which takes picture data, which includes a character group having a prescribed meaning where at least two characters are arranged in a prerilinarily determined area for one character, as the input and recognizes each character in this character group in the same manner as normal characters. CONSTITUTION:A character recognizing device 20 is provided with a pattern storage part D1 where character group paterns and standard character patterns are stored with correspondence between their features, and a recognizing part 3 retrieves the storage part D1 based on extracted features of a character group segmented by a segmenting part 1 if the character group is included in given picture data. When the pertinent character pattern is found, a intra- group character segmenting part 5 segments each character of the segmented character group, and a feature extracting part 2 and the recognizing part 3 recognize each segmented character while retrieving the storage part D1 again.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device,
Particularly, at least two or more characters are arranged in a predetermined one character area, image data including a character group having a predetermined meaning is input / output, and it is determined that this character group is different from a normal character, Furthermore, the present invention relates to a character recognition device that can recognize characters by cutting out the characters in the character group one by one.

[0002]

2. Description of the Related Art A device for recognizing characters is called a character recognizing device, which is an optical character reader (OCR) for reading printed or handwritten characters.
It is a device typified by an eader).

The OCR mainly comprises a scanning / photoelectric conversion section, a recognition processing section and an output section.

First, the material in which the character to be recognized is described is
It is sent to the scanning / photoelectric conversion unit by an operator or the like, and the scanning / photoelectric conversion unit scans this material having a two-dimensional plane in response, and the light and shade information of the character by light (1 or 0)
Is converted into an electric intensity signal and given to the recognition processing unit connected to the next stage.

The recognition processing section realizes the main part of the character recognition method. The recognition method is considered on the premise that one target character pattern exists independently, but in reality, many characters are written continuously or discretely in the material. Therefore, it is necessary to cut out a single character pattern (also called segmentation) first. This cutout is performed by printing a suitable character frame indicating the separated character frame, the position of the line where the character exists, the beginning and end of the sentence, etc. on the material to be recognized in advance, and using this. Be done.

When the character pattern is cut out as described above, pattern processing is performed such that the input is the character pattern and the output is the character name. To summarize this pattern processing, a standard pattern corresponding to each character name is stored in advance in a storage device for all the character names to be recognized, and when an input character pattern is given, it is stored accordingly. It finds the most similar one by comparing with the specified standard character pattern, and as a result, recognizes that the character name of the specified standard pattern is that of the input pattern.

The character name recognized as described above may be temporarily stored in the result storage section or the like.
When this character recognition device is used as a data input device of another computer, data processing is performed in the output section, and the resulting encoded character name is transmitted to the next computer.

The standard pattern prepared in advance is
It contains character patterns such as numbers, letters, katakana, hiragana and kanji.

[0009]

FIG. 5 is a diagram showing an example of sentence data to be recognized including a character group. As shown in FIG. 5, this text data includes character groups G1 to G3 that represent one meaning by arranging four or five characters in the print area for one kanji character. Recently, sentences including such a character group have been frequently found in newspapers and business cards. In the case of the conventional character cutting method using a character recognition device, when trying to recognize each character of the document data shown in FIG. 5, the character groups G1, G2, and G3 may be erroneously recognized as one character, or In some cases, it may be erroneously recognized as two half-width characters, and there is a problem in that information such as "co. Ltd."

Therefore, an object of the present invention is to determine at least two or more characters in a predetermined one-character area and determine a character group having a predetermined meaning, which is different from normal characters. It is an object of the present invention to provide a character recognition device that can recognize characters by cutting out characters in a character group one by one.

[0011]

A character recognition device according to the present invention comprises a first cutting means, a first characteristic extracting means, a storing means, a pattern specifying means, a second cutting means, a second characteristic extracting means, and a character. It comprises a name specifying means and an output means.

The first cut-out means inputs image data including a character group having a predetermined meaning by arranging at least two or more characters in a predetermined one-character area, and in accordance with each character area. It is configured to cut out a pattern.

The first feature extraction means is configured to extract the features of the pattern cut out by the first cutout means.

The storage means stores character patterns individually prepared for all character names to be recognized in advance and characters indicating the arrangement state of characters in a character area individually prepared for all character groups to be recognized in advance. The group pattern and the group pattern are configured to be stored in association with each other.

The pattern specifying means is configured to search the storage means based on the feature extracted by the first feature extracting means and specify a character pattern or a character group pattern having the most similar feature.

The second cutting-out means cuts out a pattern of each character of the character group cut out by the first cutting-out means in response to the pattern specified by the pattern specifying means being a character group pattern. Composed.

The second feature extraction means is configured to extract a feature for each pattern cut out by the second cutout means.

The character name specifying means is configured to search the storage means on the basis of each characteristic extracted by the second characteristic extracting means, and specify a character name corresponding to each character pattern having the most similar characteristic. It

The output means is configured to output each character name specified by the character name specifying means for each character group cut out in advance by the first cutting means.

Further, the character recognition apparatus according to the present invention is configured as described above, further comprising information addition means for outputting while adding corresponding predetermined information for each character group output by the output means. May be.

[0021]

In the character recognition device according to the present invention, image data in which at least two or more characters are arranged in a predetermined one-character area and a group of characters having a predetermined meaning and normal characters are mixed is provided. First, a storage means is provided for recognizing that a character group is different from an ordinary character by inputting, and cutting out and recognizing the characters in the determined character group one by one.

In this storage means, the pattern of the character group is
Paying attention to the fact that there is a space between each character contained in it, and that it is composed of a pattern that is quite different from a normal character pattern, while making the characteristics correspond to each pattern, And a character group pattern, which is a character group, are stored.

After the given image data is input and the character group in the image data is specified by the first cut-out means, the first feature extraction means and the pattern specifying means in response, the specified character group is Each character pattern is cut out in the character group by the two cutting-out means, the characteristics are extracted by the second characteristic extracting means for each cut-out character pattern, and the character name is specified based on the extracted characteristics. The means searches the storage means and specifies the corresponding character name. Then, since the output means outputs each specified character name for each character group, it is possible to obtain the meaning of the character group to be output as each character name in the same manner as a normal character.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a schematic block diagram of a character recognition device according to an embodiment of the present invention. The character recognition device 20 inputs, for example, the document data to be recognized shown in FIG. 5 as image data, and performs character recognition processing on this. At this time, normal characters are output as recognized character names, and the character groups G1 and G2 and "Principal Researcher" are displayed in which the four characters "corporation" are collectively displayed in the normal character area. The character group G3 to be displayed is configured so that each character in the character group is individually recognized and the meaning displayed by each character group is represented by a character name that is continuously output.

The character recognition device 20 includes a cutout unit 1, a feature extraction unit 2, a recognition unit 3, a character group determination unit 4, a character group cutout unit 5,
A recognition standard pattern storage unit D that is a kind of storage device that includes the item recognition unit 6 and the result output unit 7 and that is used by the recognition unit 3 to retrieve and read the data stored therein.
1. An item determination database D2 that is accessed by the item recognition unit 6 and whose data is read out.

FIG. 2 is a diagram for explaining the data storage format of the recognition standard pattern storage unit D1 according to one embodiment of the present invention.

In FIG. 2, the recognition standard pattern storage unit D1
Are pre-classified into normal character categories and character group categories. In the normal character category, data relating to a character standard pattern is stored, and the data is stored at addresses 0 to (N-1). Further, in the character group category, data relating to the character group standard pattern is stored, and these data are stored at addresses N to N + 6.

As shown in the figure, the recognition standard pattern storage section D1 stores data relating to each character standard pattern and each character group standard pattern in a record format, and each record stores an address, a character pattern and a character group. Pattern I1 and density distribution characteristic data I2
Is included as an item.

The density distribution characteristic data I2 stores data representing the status of the density distribution of an image in a predetermined one-character area (a full-width character area in this embodiment) of the corresponding character pattern and character group pattern I1. It

In the area where the character group standard pattern is stored, the density distribution characteristic data I2 is stored as data relating to the distribution of image density of the corresponding character pattern and the pattern stored in the character group pattern I1. .. The data I2 stored in this character group standard pattern is also stored together with the characteristic data that there is a space between each character in the character group.

Next, patterns 1 to 7 of the character group standard patterns stored in the recognition standard pattern storage unit D1.
Will be described.

FIGS. 3A to 3G are views for explaining a character group standard pattern according to an embodiment of the present invention.

As shown in the figure, when the character group is cut out as full-width characters, five patterns 1 to 5 shown in FIGS. 3A to 3E occur. When the character group is cut out as half-width characters as shown in FIGS. 3A and 3G, two patterns, pattern 6 and pattern 7, are generated. It can be said that the character group patterns of the patterns 1 to 7 are data representing the arrangement state of each character in one character group.

Returning to FIG. 1, the cutout unit 1 inputs the image data, and cuts out a character included in the image data, for example, a character pattern for each character area of full-width characters.

The feature extraction unit 2 performs feature extraction (density distribution feature) of characters for a given character pattern.

The recognition unit 3 accesses the recognition standard pattern storage unit D1 for each of the cut out character patterns based on the characteristics extracted in advance, and performs matching with all the character standard patterns and the character group standard patterns. , And operates to identify and read a standard pattern having the most similar features.

The character group determination unit 4 determines whether the standard pattern read by the recognition unit 3 is a character group standard pattern. This determination is performed based on the address value at which the recognition unit 3 performs the matching process and the standard pattern having the most similar feature is stored. As can be seen from FIG. 2, the address value obtained as a result of the matching is
If the address is N or more, the extracted character pattern is
It is determined to be a character group pattern. If the character group pattern is determined in this way, the process proceeds to individual recognition processing in order to recognize each character in the determined character group.

The character group cut-out unit 5 takes a histogram representing horizontal and vertical image density levels only for the image of the character group determined to be a character group pattern by the character group determination unit 4. And based on this histogram,
Cuts out each character in the character group. Each character cut out from the character group in this way has a character standard pattern.

Each character pattern cut out by the character group cutout unit 5 is processed in the feature extraction unit 2 and the recognition unit 3 in the same manner as described above. Specifically, the recognition standard pattern storage unit D1 is accessed based on each extracted feature, and the matching process is performed only for the character standard pattern, that is, the character standard pattern of address 0 to address (N-1). As a result, a recognition result is obtained for each character in the character group. The recognition result passes through the character group determination unit 4 and is given to the item recognition unit 6.

When the recognizing unit 3 recognizes, for example, the character group G1 shown in FIG. 5 as "stock company", the item recognizing unit 6 recognizes "stock", "expression", "kai", and "company". It is configured to output sequentially.

Further, the item recognition section 6 operates to search the item determination database D2 as necessary, and add the information obtained by this search result to the data to be output and output it. For example, when the character recognition device 20 recognizes a character from image data printed on a business card, the item determination database D2 sets the item of the line in which the character group G1 of FIG. This is a database that is accessed to determine that the item in the line in which the character group G3 in FIG. 5 is written is "title".

The result output unit 7 has a function of converting the recognized character name output by the item recognition unit 6 and the item read from the item determination database D2 into a corresponding code and outputting the code to the outside of the apparatus.

FIG. 4 is a processing flow chart showing the operation of the character recognition device 20 shown in FIG.

Next, with reference to FIGS. 1 to 5, the character recognition operation of the apparatus 20 will be described according to the processing flow of FIG.

In the recognition standard pattern storage section D1,
Character pattern I for all the character names you want to recognize in advance
1 and its density distribution characteristic data I2 are stored, and the character group pattern I1 and corresponding density distribution characteristic data I2 are stored for all the character groups to be recognized in advance.
Assumed to be stored.

Further, in the item judging database D2,
It is assumed that the data required for recognizing the item in the item recognizing unit 6 is stored in advance.

The character recognition device 20 responds to, for example, the optically read image data of the light and shade of the document data to be recognized as shown in FIG.
Execution of the processes after step ST1 shown in FIG. 4 is started.

First, in the processing of step ST1, the cutout unit 1 inputs the given grayscale image data, and performs line extraction and character cutout.

In the line extraction, a histogram representing the degree of shading of the image in the horizontal direction or the vertical direction is taken based on the given image data, and a portion having a small degree of shading is judged to be the space between lines based on this histogram, and accordingly, Extract one line. At this time, it is also determined at the same time whether the supplied image data is vertically written image data or horizontally written image data.

When the line extraction is performed in this way, character cutting is next performed. For character extraction, a histogram that represents the degree of shading of an image in the vertical direction (in the case of vertical writing) is taken for the extracted line, and a character pattern that seems to be one character is extracted from the state of density variation based on the histogram. Do. The extraction of the character pattern is performed for each full-width character or each half-width character area.

When the character pattern is extracted as described above, the processing relating to recognition is performed in the processing of the next step ST4. This processing is performed by the feature extraction unit 2 and the recognition unit 3.

The feature extraction unit 2 extracts the image density distribution feature of the pattern for each predetermined character region cut out by the cutout unit 1 as described above, and the extracted feature is given to the recognition unit 3. .. In addition, when the pattern provided as shown in FIG. 3 is a character group pattern, the feature extraction unit 2 also provides the recognition unit 3 with the feature extraction data regarding the existence of a blank space between each character.

The recognition section 3 accesses the recognition standard pattern storage section D1 based on the given feature data and performs matching processing. As a result of this matching processing, when the standard character pattern or the character group pattern I1 having the most similar features is specified, the specified character pattern or character group pattern I1 and the address data at which it is stored. It is given to the character group determination unit 4.

The character group determination unit 4 executes the process of step ST5 based on the data of the given address. In the process of step ST5, the character group determination unit 4 determines whether the cut-out pattern is a character standard pattern or a character group standard pattern, depending on whether the given address value is smaller than N or not. At this time, if the given address value is N or more, the character group determination unit 4 determines that the cut-out character pattern is a character group standard pattern,
Accordingly, the process proceeds to step ST6 and subsequent steps shown next.
On the other hand, if it is determined that the character group pattern is not a character group pattern, the process proceeds to step ST8 and subsequent steps, which will be described later.

In the process of step ST6, pattern cutting is performed for each character in the cut-out character group pattern. The pattern cutout of each character is performed in the character group cutout unit 5.

When each character in the character group is cut out into a character pattern by the cutout unit 5, the feature extraction unit 2 extracts the density distribution feature for each character pattern as described above in the process of step ST7. The recognizing unit 3 accesses the character standard pattern in the recognition standard pattern storage unit D1 based on the extracted features and identifies the corresponding character pattern I1. Specifically, the character group cutout unit 5 takes a histogram representing the degree of grayscale of the image in the horizontal and vertical directions only for the image of the character group pattern cut out by the cutout unit 1, and based on this histogram After extracting each character pattern, feature extraction is performed for each character pattern. Then, the character standard pattern stored in the storage unit D1 is accessed based on the extracted feature to perform matching processing. As a result, the recognition result for each character in the character group is obtained as a character name. The character name which is the recognition result obtained in the character group pattern and the normal character pattern is given to the item recognition unit 6.

The item recognition section 6 executes the process of step ST8 in FIG. Item determination is performed in the process of step ST8. For example, when the character group G1 in FIG. 5 is recognized, the item recognition unit 6 recognizes “share”, “expression”,
The result is output to the result output unit 7 sequentially and continuously as “kai” and “company”.

When the character recognition device 20 inputs the image data printed on the business card, the character group printed on the business card, for example, the character groups G2 and G in FIG.
When 3 is recognized, the item determination database D2 is searched based on the recognition result, the corresponding item is read, and the read item is output together with each character name of the recognized character group. For example, when the character group G2 in FIG. 5 is recognized, the line of the character group G2 is specified by the item "company name",
When the character group G3 is recognized, the line of the character group G3 is processed to be recognized by the item "name". Item recognition unit 6
The item recognition result and the recognized character name output by are output to the result output unit 7.

The result output unit 7 performs the output-related processing in step ST9. Result output unit 7
Operates to print out the derived item recognition result and the character name obtained by the recognition, output an image, output an audio, or output a code.

When a computer is connected to the character recognition device 20 and the device 20 itself is an input device of the computer, the result output unit 7 is recognized by the computer so that the data processing can be easily performed. The character name and the item recognition result are converted into a predetermined code and output.

The processing relating to one pattern cut out in the processing of step ST3 has been described above.
When a row has a plurality of patterns, the processes in steps ST3 to ST9 are repeatedly executed by the number of patterns included in the row. Further, when the given image data has a plurality of lines, step ST2
The process of step ST9 is further repeated by the number of lines included in this image data while the extraction of the character pattern of each line in step ST3 is repeated.

When the information of the character group to be recognized is constant (for example, "corporation", "foundation", etc.) or when it is known in advance, the knowledge is used to make the above-mentioned characters. A function for confirming whether the recognition result of the group is correct or not and automatically correcting if the recognition result is incorrect may be added.

Further, since the arrangement of each character in the character group is also different between vertical writing and horizontal writing, vertical writing is performed based on the determination of vertical writing and horizontal writing accompanying the line cutting result obtained in the process of step ST2 of FIG. Alternatively, by referring to a predetermined rule of horizontal writing, the order in outputting the character names may be performed based on the recognition result of each character in the character group.

[0065]

As described above, according to the present invention, image data including a character group having a predetermined meaning in which at least two characters are arranged in a predetermined one-character area is input. Also, this character group can be recognized, each character in the character group can be recognized individually, and each character name that constitutes the character group can be output for each character group. There is an effect that the meaning represented by the character group can be recognized exactly as in the normal character.

Further, since the information addition / output means outputs while adding the information in which the corresponding predetermined meanings are classified in advance for each character group to be output, using this output result, the added information can be regarded as an important key word. Then, there is an effect that the data consisting of the character names that follow can be classified into files and managed.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a character recognition device according to an embodiment of the present invention.

FIG. 2 is a diagram for explaining a data storage format of a recognition standard pattern storage unit according to an embodiment of the present invention.

FIGS. 3A to 3G are views for explaining a character group standard pattern according to an embodiment of the present invention.

FIG. 4 is a process flow chart showing an operation of the character recognition device shown in FIG. 1.

FIG. 5 is a diagram showing an example of document data to be recognized including a character group.

[Explanation of symbols]

 1 cutout part 2 feature extraction part 3 recognition part 4 character group determination part 5 character group cutout part 6 item recognition part 7 result output part D1 recognition standard pattern storage part D2 item determination database Indicates the same or a corresponding portion.

Claims (2)

[Claims] 1. At least two or more characters are arranged in a predetermined character area, image data including a character group having a predetermined meaning is input, and a pattern is cut out for each of the character areas accordingly. First cut-out means, first feature extraction means for extracting the features of the pattern cut out by the first cut-out means, and character patterns individually prepared for all character names to be recognized in advance, A storage unit that stores a character group pattern indicating the arrangement state of characters in the character area, which is individually prepared for all the character groups to be recognized in advance, in association with each of the characteristics, and the first characteristic. Pattern specifying means for searching the storage means based on the characteristics extracted by the extraction means, and specifying the character pattern or the character group pattern having the most similar characteristics; Second cutting means for cutting out a pattern of each character of the character group cut out by the first cutting means in response to the pattern specified by the turn specifying means being the character group pattern; A second feature extraction unit that extracts the feature for each of the patterns cut out by the second cutout unit, and a search for the storage unit based on each feature extracted by the second feature extraction unit. A character name specifying unit that specifies the character name corresponding to each of the character patterns having the most similar characteristics, and each of the character names specified by the character name specifying unit,
A character recognition device, comprising: an output unit that outputs each of the character groups cut out by the first cutting unit. 2. The character recognition device according to claim 1, wherein the output means further comprises an information addition output means for outputting while adding predetermined information corresponding to each of the output character groups.